KR20100038455A - Partially complex modulated filter bank - Google Patents

Abstract

The present invention relates to an apparatus for processing a plurality of real-valued subband signals comprising a first real-valued subband signal and a second real-valued subband signal to provide at least a complex-valued subband signal with a multiband filter for providing an intermediate real-valued subband signal and a calculator for providing the complex-valued subband signal by combining a real-valued subband signal from the plurality of real-valued subband signals and the intermediate subband signal.

Description

Partial Complex Modulation Filter Bank {PARTIALLY COMPLEX MODULATED FILTER BANK}

The present invention relates to an apparatus and method for processing a plurality of complex subband signals, and more particularly, to an apparatus and method for processing a plurality of real value subband signals in the field of encoding and decoding audio signals.

It is noted that a bank of complex exponential modulation filters is an efficient tool for adjusting the spectral envelope of an audio signal. The paper by P. Ekstrand, "Bandwidth extension of audio signals by spectral band replication", Proc. 1st IEEE Benelux Workshop on Model based Processing and Coding of Audio ( MPCA-2002), pp. 53-58, Leuven, Belgium, 2002). One application of this property is audio coding based on spectral band replication (SBR). Another useful application is a complex filter bank that includes frequency selective panning and spatialization for parametric stereo. E. Schuijers, J. J. Breebart, H. H. Purnhagen, J. J. Engdegard's paper, "low complexity parametric stereo coding" (Proc. 116th AES convention, 2004, paper 6073). Another useful application is parametric multichannel coding, which is J. J. Herre et al., "Reference model architecture for MPEG spatial audio coding", Proc. 118th AES convention, 2005, paper 6447. In these applications, the frequency resolution of the complex filter bank is further improved in the low frequency region by the sub-subband filter. This allows the combined hybrid bank to obtain a frequency resolution that allows processing of spatial cues at a spectral resolution close to the spectral resolution of the stereoacoustic listening system. Since this additional filtering does not cause aliasing by itself, no matter what modifications are made, the performance of the hybrid filter bank is determined by the aliasing of the first filter bank.

If the constraints on computational complexity make it difficult to use complex exponentially modulated filter banks and only allow the implementation of cosine modulation (real values), then serious aliasing will occur when using this filter bank for spectral envelope adjustment. Five. O. Shamida et al., "Low power SBR algorithm for the MPEG-4 audio standard and its DSP implementation", Proc. 116th AES As disclosed in Convention, 2004, paper 6048, such aliasing can be relaxed to some extent by adaptive subband gain grouping (or gain locking). However, this method can achieve an optimal action only when only the high frequency components of the signal need to be modified. For the purpose of panning in parametric multichannel coding, the amount of gain locking required to avoid hearing in the low frequency region greatly reduces the frequency selectivity of the two filter banks and actually reduces the additional frequency selectivity of the hybrid filter bank. Makes it impossible to achieve. As a result, the sound impression becomes narrower and causes problems in correct sound source arrangement. If complex signal processing is maintained only in the low frequency region, which is more important in perceptual terms, a much better compromise between performance and complexity will be achieved.

It is an object of the present invention to provide a more efficient technical idea for providing a signal enabling an operation with improved performance, and a more efficient technical idea for providing a signal with less distortion.

The object of the invention can be achieved by an apparatus according to claim 1, a system according to claim 16, a method according to claim 18, and a computer program according to claim 19.

According to one aspect of the invention, there is provided an apparatus for processing a plurality of real valued subband signals including a first real valued subband signal and a second real valued subband signal to obtain a complex subband signal, comprising: Filter the first real subband signal to obtain a filtered subband signal, filter the second real subband signal to obtain a second filtered subband signal, and obtain a real intermediate subband signal A multi-band filter for combining the first filtered subband signal and the second filtered subband signal to supply the real value intermediate subband signal; And combining the real value subband signal from the plurality of real value subband signals as a real part of the complex subband signal with a signal based on the intermediate subband signal as an imaginary part of the complex subband signal. There is provided an apparatus including a calculation unit for providing the complex subband signal.

According to a second aspect of the present invention, there is provided an apparatus for processing a plurality of complex subband signals each including a first complex subband signal and a second complex subband signal to obtain a real value subband signal. Extracting a first imaginary part from the first complex subband signal, and extracting a second imaginary part from the second complex subband signal, wherein the first, second, or An extractor for extracting a real part from a third complex subband signal; Filter the first imaginary part to obtain a first filtered imaginary part signal, filter the second imaginary part to obtain a second filtered imaginary part signal, and filter the first imaginary part to obtain an intermediate subband signal A multi-band filter for combining the signal and the second filtered imaginary signal to provide the intermediate subband signal; And a calculator configured to combine the real part and the intermediate subband signal to provide the real value subband signal.

The present invention can be processed to provide at least one complex subband signal that allows a plurality of real valued subband signals to operate with better performance than manipulation of the plurality of real valued subband signals. The computational complexity of the processing of the real value subband signal is increased slightly. More specifically, the present invention relates to a multiband filter to obtain a complex subband signal that can be easily manipulated without minimal aliasing and significant amount of distortion as compared to directly manipulating multiple real value subband signals. The calculation is made based on the fact that multiple real value subband signals can be processed.

In one embodiment of the present invention, a characteristic apparatus of the present invention for processing a plurality of real valued subband signals is disclosed, which provides a plurality of complex subband signals from a subset of the plurality of real valued subband signals, The second subset of the plurality of real value subband signals is provided as another plurality of real value subband signals without being processed into a corresponding number of complex subband signals. Thus, this embodiment discloses a partial complex modulation analysis filter bank, where the complex subband signal is complex exponential modulation in terms of the evaluation of energy stability and minimal aliasing resulting from linear time invariant deformations such as level adjustment and filtering. It has the same advantages as the corresponding subband signal from the filter bank. In addition, as a further advantage, the computational complexity is significantly reduced compared to complex filter banks that process complex valued signals.

As will be described later, in another embodiment of the present invention, it may further include a deformation-transformer for implementing the time-varying operation and / or non-linear operation. Examples of such embodiments are high performance SBR, application of spatial parameters, and other applications. In these embodiments, all the advantages of the manipulator of the corresponding complex bank are shown in the complex part of the partial complex filter bank of the embodiment of the present invention.

In another embodiment of the present invention, the complex number subband signal outputted by the characteristic device of the present invention is a plurality of other real value subband signals passed by the characteristic device of the present invention for processing the multiple real value subband signals. To ensure time synchronization with the signal, it is delayed by a delay.

A second aspect of the present invention uses an extractor to extract the real imaginary part of at least two complex subband signals from at least two complex subband signals and extract a first, second, or third complex subband signal. Extracts the real part from the multi-band filter, provides an intermediate signal based on the imaginary part, and uses a calculation unit to combine the real part signal and the intermediate signal to provide a real value subband signal. It is based on the fact that the Sochi subband signal can be reduced more efficiently to a real value subband signal with less distortion and minimal aliasing. More specifically, the present invention provides that, before performing the optional real synthesis, another multiband filter converts the complex subband signal back to the real value subband signal, and the overall performance of the reproduction and signal processing is complex. On the basis of comparable performance of the filter bank.

According to a specific embodiment of the present invention, the extractor may be implemented as a separator, for example if two or more real valued subband signals are provided. In this case, it is useful to perform a process of extracting their proper real and imaginary parts from all complex subband signals.

Conversely, if one real subband signal is obtained based on three or more different complex subband signals, the extractor may be implemented as a separator that separates each complex subband signal into a real part and an imaginary part. . In this case, the imaginary part signal and the real part signal which are not required for this further processing can be ignored. Thus, the terms "separator" and "extractor" may be used synonymously in the present invention.

In addition, in the present invention, the imaginary part and the imaginary part signal as well as the real part and real part signals may also mean all signals having values corresponding to the imaginary part or the real part of the value of the complex subband signal. Further, in principle, the imaginary part signal, the real part signal, or both may be a real value or a complex value.

In one embodiment of the present invention, a characteristic apparatus of the present invention for processing a plurality of complex subband signals is also provided with a plurality of real value subband signals, the plurality of complex subband signals as described above. The multiple real subband signals are processed and provided in an unfiltered form at the output of the device. Thus, this embodiment constitutes a partial complex modulation synthesis filter bank. The main advantage of this embodiment is that the overall performance of the reproduction and signal processing is comparable to the performance of the complex filter banks for multiple complex subband signals, and real filters in the remaining frequency ranges represented by the multiple real subband signals. It is comparable to the performance of the bank. An additional advantage of this embodiment is that the computational complexity is only slightly increased compared to the real value filter bank. A further advantage of this embodiment is that a seamless transition between the two frequency ranges represented by the multiple complex subband signals and the multiple real subband signals is implemented in a particular boundary band processing. In addition, an additional advantage of this embodiment is that the computational complexity is significantly reduced compared to the multiple filter banks that process complex signal.

In another embodiment of the present invention, a characteristic device of the present invention for processing a plurality of real valued subband signals and a characteristic device of the present invention for processing a plurality of complex valued subband signals are provided. An apparatus also provides a system for passing many other real subband signals. Between these two devices characteristic of the present invention, the first manipulator and the second manipulator are arranged in a number of complex valued subband signals and a plurality of other complex valued subband signals output by the characteristic device of the invention. Transform each real-valued subband signal. The first manipulator and the second manipulator can perform linear time invariant deformations such as envelope adjustment or filtering. Thus, in the above system, the performance of the reproduction and signal processing is comparable to the performance of the complex filter bank in the frequency range represented by the multiple complex subband signals and represented by the other multiple real value subband signals. The frequency range is comparable to the performance of the real filter bank, and the computational complexity is slightly increased compared to the case of directly transforming a large number of real-valued subband signals, but the signal can be manipulated with better performance. As outlined above and in detail below, the manipulators in other embodiments are not limited to linear and / or time invariant manipulators.

In another embodiment of the characteristic device of the present invention for processing a plurality of complex subband signals, another multiple real valued subband signal is provided in the characteristic device of the present invention for processing a plurality of complex subband signals. In order to secure time synchronization for the output real value subband signal, it is passed in a delayed form using a delay.

The present invention will be described through exemplary embodiments with reference to the accompanying drawings, which embodiments do not limit the scope or spirit of the invention. Preferred embodiments of the present invention are described with reference to the following drawings.
1 illustrates partial complex signal processing;
2 illustrates a partial complex analysis filter bank;
3 shows a partial complex synthesis filter bank;
4 illustrates multi-band filtering;
5 shows a spectrum of a raw signal comprising a number of sinusoidal components;
6 shows the spectrum of a signal obtained by performing analysis and synthesis without subband modification in a partial complex filter bank that does not employ a seamless transition feature as disclosed herein;
7 shows the spectrum of a signal obtained by modifying the subband domain of a complex filter bank;
8 shows the spectrum of a signal obtained by modifying the subband domain of a real filter bank;
9 shows a spectrum of a signal obtained by modifying the subband domain of a partial complex filter bank as disclosed herein;
10 illustrates a hybrid QMF analysis bank for time / frequency transform in spatial audio coding;
11 shows a hybrid QMF synthesis bank for time / frequency transform in spatial audio coding;
12 is a flowchart of a real value analysis QMF bank;
Figure 13 illustrates one embodiment of an apparatus for processing with a plurality of real-valued subband signals real-complex converters in accordance with the present invention; And
14 is a diagram illustrating one embodiment of an apparatus for processing multiple complex subband signals with a complex-real number converter in accordance with the present invention.

An embodiment for exemplarily describing the principle of a partial complex modulation filter bank according to the present invention is disclosed. It will be understood by those skilled in the art that the configurations described herein and modifications thereof may be used and variations are possible. It is intended that the invention be limited only by the scope of the claims of the present invention and not by the specific details in the detailed description.

1 shows the principle of partial complex signal processing based on a partial complex analysis filter bank 101 and a partial complex synthesis filter bank 104. The digital audio input signal is applied to the partial complex analysis filter bank 101. Of the total L subband signals, the analysis bank outputs K complex subband signals and (LK) real subband signals, where K and L are positive integers and have a relationship K≤L. . The first modification process 102 is performed on the real valued subband signal, and the second modification process 103 is performed on a complex valued subband signal. The purpose of these transformation processes is to shape the audio signal in the time and frequency domain. Subsequently, the modulated subband signals are applied to a partial complex synthesis filter bank 104 which produces the processed digital audio signal as an output.

2 illustrates the components of one embodiment of a partial complex analysis filter bank 101 as disclosed herein. The digital audio input signal is analyzed by an L-band cosine modulation filter bank 201 that splits L real value subband signals into two groups at its output. A first group comprising K real subband signals is filtered by multiband filter 204, the output of which is multiplied by a negative imaginary unit by multiplier 205 and delayed by delay 203. The K real value subband signals are added to the K real value subband signals to obtain K complex value subband signals. The subband signal is adjusted by a fixed real gain adjuster 207, which is output as K multiple subband signals of partial complex analysis. A second group comprising (L-K) real subband signals is supplied to a delay 202 having an output constituting the real subband of the partial complex analysis. The amount of delay in delays 202 and 203 is adjusted to compensate for the delay introduced by multiband filter 204. The delayer 202, delayer 203, multiband filter 204, multiplier 205, adder 206, and fixed real gain adjuster 207 constitute a real-complex converter 210. Is supplied with K real value subband signals, and is also supplied with K complex value subband signals and (LK) real value subband signals providing (LK) real value subband signals. The multiplier 205 and the adder 206 constitute a calculator 215, which is based on at least one real value subband signal as a real part signal of the complex subband signal and is a complex subband signal. Supply at least one complex subband signal based on the at least one real value subband signal as an imaginary part signal of.

3 illustrates one embodiment of a partial complex synthesis filter bank 104 in accordance with the present invention. The (L-K) real valued subband signals are delayed by delay 304 and fed to the (L-K) inputs of L-band cosine modulated synthesis filter bank 308. The K complex subband signals have their first gain adjusted by a fixed real gain adjuster 301. The real part and the imaginary part of the complex subband signal are extracted by the real part extractor 302 and the imaginary part extractor 303, respectively. The imaginary part of the complex subband signal is filtered by the multiband filter 306 and its output is added by the adder 307 to the real part of the subband signal delayed by the delay 305. The amount of delay in delay 304 and delay 305 is adjusted to compensate for the delay introduced by multiband filter 306. The output of adder 307 is fed to the remaining K inputs of L-band cosine modulated synthesis filter bank 308. The real part extractor 302, together with the imaginary part extractor 303, constitutes a separator 309 for separating the complex subband signal into a real value real part signal and a real value imaginary part signal. More specifically, the real part extractor 302 provides a real part signal, and the imaginary part extractor 303 provides an imaginary part signal. In the embodiment shown in FIG. 3, the separator 309 separates or processes the K complex subband signals into K real value imaginary part signals and K real value imaginary part signals.

However, as described above, the separator 309 may be implemented as an extractor. In this case, a configuration for separating all complex subband signals into a real part signal and an imaginary part signal is not adopted. Thus, the separator 309 is also used in the same sense as the extractor 309 for extracting the real part signal and the imaginary part signal from the complex subband signal.

Separator 309 comprising a fixed real gain adjuster 301, a real part extractor 302 and an imaginary part extractor 303, a retarder 304, a retarder 305, a multiband filter 306, and an adder. 307 constitutes a complex-real converter 310 which is a feature of the present invention. The complex-real number converter 310 converts K complex subband signals into K real value subband signals, and (LK) real number subband signals in a delayed form at the output terminal of the complex-real number converter 310. To feed.

에 의해 각각 (q(m)+p(m)+1)개의 입력 (m-q(m)),..., m,..., (n+p(m))을 필터링하고 그 결과를 구성요소(403m)에서 합산함으로써, 생성된다. 제약 조건 (m-q(m))≥0과 (m+p(m))≤(K-1)은 유지되어야 한다. 상술한 바와 같이, 본 발명에 따르면, 다음과 같은 낮은 계산 복잡도를 가진 다중 대역 필터(204 및 306)를 이용하여 복소 표현을 얻는 방법이 개시된다.4 takes K real value subband signals as input (0,1,2, ..., (K-1)) and outputs K real value subband signals (0,1,2, ...). (K-1)) shows the operation of the multi-band filter 401. In the linear system representation, this is simplified to a linear time invariant discrete time multiple input multiple output (MIMO) system. mth output, filter on component 402m

Filter each of (q (m) + p (m) +1) inputs (mq (m)), ..., m, ..., (n + p (m)) and construct the result By summing in element 403m. Constraints (mq (m)) ≥0 and (m + p (m)) ≤ (K-1) must be maintained. As described above, according to the present invention, a method of obtaining a complex representation using multiband filters 204 and 306 having the following low computational complexity is disclosed.

(1)

(One)

(2)

(2)

Further, the filter F _{m, - 1} and the filter F _m, the degree of similarity of _one may be utilized to further reduce the computational complexity. Particularly small values of q (m) and p (m), represented by equations (1) and (2), can be used when the standard filter of the cosine modulated filter bank has a sufficiently high stop band attenuation. . This implicit condition requires the minimum length of a standard filter. For shorter standard filters, the q (m) and p (m) values must be increased. However, since the length of the filters F _{m , r} is proportional to the length of the standard filter, the method according to the invention still maintains the computational efficiency.

The filter implemented in the multi-band filter 401 may in principle be any kind of filter with all filter characteristics. In the embodiment shown in FIG. 4, a multiband filter F _{m , 0} that maps a subband signal with subband index m to a subband signal with the same subband index _{m , 0} typically has a center frequency at π / 2. It is a band pass filter. For a multiband filter that combines three subband signals into one subband signal as a filter bank signal, the other two multiband filters F _{m , -q (m)} and F _{m , + p (m)} are typical As a high pass filter to a low pass filter, the exact type of these filters depends on the subband index m. The corresponding multiband filter, where a multiband filter 401 is employed to combine more than three subband signals to obtain a filter subband signal that is not a "boundary" subband signal, m. The type of may be a band pass filter, a high pass filter, a low pass filter, a band cut filter, or a pass filter.

1 to 3 illustrate a method for transforming a discrete time audio signal.

Filtering the discrete time audio signal by a cosine modulation analysis filter bank;

Generating a complex subband sample for the subband subset by multiband filtering;

Transforming the real subband sample and the complex subband sample;

Converting the complex sample into a real sample by multiband filtering; And

Disclosing the real subband sample through a cosine modulated synthesis filter bank.

5 shows a portion of a magnitude spectrum of an original signal comprising a number of sinusoidal components. This spectrum is obtained using a windowed discrete Fourier transform. The frequency index n is normalized so that the frequency index n corresponds to the frequency nπ / L, where L = 64. The sampling frequency of the digital audio signal is f _s Then, the frequency range shown in Fig. 5 is from (5/64) f _s / 2 to (11/64) f _s / 2. In the normalization process, the subband of frequency index n of a complex or real modulated filter bank having L subbands is located at the center of its main lobe in the range from frequency index n to frequency index (n + 1). Has a response. This rule remains the same in FIGS. 5-9.

In other words, each subband to subband signal is associated with an exponent n or exponent m and the center frequency of the corresponding subband. Subband signals to subbands are configured according to the center frequency associated with the subband signal, for example so that a larger exponent corresponds to a higher frequency.

Figure 6 shows the spectrum of a signal obtained by analysis and synthesis without subband modification in a partial complex filter bank that does not employ a seamless transition according to the present invention. In particular, as a more basic configuration, the partial complex analysis filter bank 101 is composed of two filter banks having subbands in which L = 64, the first bank is a complex exponential modulation type, and the second bank is cosine modulation. It is considered to be a brother. These filter banks exhibit almost complete reproducibility even when used individually. In the configuration considered here, the first subband having K = 8 from the first complex bank and the remaining subbands from the second real bank are adopted. The input signal is the same as the signal shown in FIG. 5, and compared with FIG. 5, it can be seen that the aliasing component is introduced near the frequency index 8 representing the transition frequency between the majority subband and the real subband. Neglecting that the complexity of this basic configuration is higher than that of a simple complex bank, these examples need special handling of the transition between the complex subband and the real subband. If no deformation is performed at all in the first modification process 102 and the second modification process 103, the digital audio output from the complex synthesis filter bank 104 should be generated. This output is sensibly indistinguishable from the output to the partial complex analysis filter bank 101. The partial complex analysis filter bank and the partial complex synthesis filter bank according to the present invention shown in Figs. 2 and 3 have the same configuration. In particular, the corresponding magnitude spectrum of the processed signal is the same as in FIG. 5. Thus, the chain structure of a multiband analysis filter, an analysis filter bank and a synthetic multiband filter, or a synthesis filter bank, i.e., the chain structure of the multiband analysis filtering and the multiband synthesis filtering, has an approximate perfect reproducibility, e.g., a code change. Can be represented.

을 지수 n을 가진 부대역에 인가하는 것을 포함한다. 여기서, 이득

은 n에 대한 감소함수이다. 도 5와 비교하여, 정현파 성분은 단순하게 변화하는 크기를 가진다. 이는 원시 신호의 포락선 조절 내지 등화(equalization)의 원하는 형태를 설명하고 있다. 실수 코사인 변조 필터 뱅크로써 동일한 변형을 수행하면, 도 8에 도시된 주파수 분석에 따른 출력 신호를 얻을 수 있다. 추가적으로 앨리어싱된 정현파 성분은 도 7에 도시된 바와 같은 원하는 형태와 상당히 다른 결과를 얻게 되고, 그 왜곡은 청취가능하게 된다. 각각의 개별적 필터에 대해 11개의 필터 탭으로 도 4에 도시된 바와 같이 다중 대역 필터에 의해 구현되고 도 2 및 도 3에 도시된 부분 복소 필터 뱅크에서 동일한 이득 변형을 적용하면, 도 9의 크기 스펙트럼이 얻어진다. 다시, K=8이 선택되면, 도시된 바와 같이, 복소 필터 뱅크 처리(도 7 참조)의 성능이 주파수 지수(K-0.5 = 7.5) 이하이고, 실수 필터 뱅크 처리(도 8 참조)의 성능이 이 주파수 지수 이상인 것을 알 수 있다.7 shows the spectrum of a signal obtained by the modification in the subband domain of a complex exponential modulation filter bank. Transformation is a gain

Is applied to the subband with index n. Where the gain

Is the decreasing function for n. In comparison with FIG. 5, the sinusoidal component has a simply varying magnitude. This illustrates the desired form of envelope regulation or equalization of the primitive signal. By performing the same transformation with a real cosine modulated filter bank, an output signal according to the frequency analysis shown in FIG. 8 can be obtained. In addition, the aliased sinusoidal component results in significantly different results than the desired shape as shown in FIG. 7, and the distortion becomes audible. With the eleven filter taps for each individual filter implemented by a multi-band filter as shown in FIG. 4 and applying the same gain variant in the partial complex filter banks shown in FIGS. 2 and 3, the magnitude spectrum of FIG. Is obtained. Again, when K = 8 is selected, as shown, the performance of the complex filter bank processing (see Fig. 7) is below the frequency index (K-0.5 = 7.5), and the performance of the real filter bank processing (see Fig. 8) is It turns out that it is more than this frequency index.

Accordingly, the system according to the present invention includes frequency selective spatialization, frequency selective panning, spectral envelope adjustment, or equalization of an audio signal using downsampled real value subband filter banks. Accordingly, aliasing in the selected frequency range can be suppressed by converting the corresponding subset of subband signals into a complex subband signal. Under the assumption that aliasing outside the selected frequency range is hardly detected or otherwise mitigated, the present invention has the effect of greatly reducing the amount of calculation compared to the case of using a complex filter bank.

Modulation Filter Bank

To facilitate the calculation, a majority exponentially modulated L-band filter bank is modeled by continuous time windowed transform using the following composite waveform.

(3)

(3)

Where n and k are integers with n≥0

(4)

(4)

The window function v (t) is assumed to be a real function. When the waveform e _n (t) = C _n (t) + is _n (t) is separated into a real part and an imaginary part, a synthesized waveform for a cosine modulated filter bank and a sine modulated filter bank is obtained.

(5)

(5)

The result of a filter bank with L subbands for discrete time signals is obtained by appropriate sampling of the t-variable at intervals 1 / L. The dot product between signals is defined as follows.

(6)

(6)

Here, the asterisk means a conjugate (pair) complex number. For discrete trial signals, the integral is replaced by the summation. The calculation of cosine modulation filter bank analysis and sine modulation filter bank analysis of signal x (t) is expressed as follows.

(7)

(7)

에 대해, 대응하는 합성 연산은 다음과 같다.Given subband signal

For, the corresponding synthesis operation is as follows.

(8)

(8)

에 대한 완전한 재현성

을 얻을 수 있도록 윈도우 함수 ν(t)가 설계될 수 있다는 것이 잘 알려져 있다. 근사적 완전 재현성의 설계에서는, 이러한 동등성이 근사성으로 대체될 것이다.For discrete time signals, the summation for subband index n is limited to (L-1). According to the theory of cosine / sine modulation filter banks and lap transforms, any undeformed subband signal by a combination of analytic and synthesis operations

Full reproducibility for

It is well known that the window function v (t) can be designed to obtain. In the design of approximate perfect reproducibility, this equivalence will be replaced by approximation.

The operation of the complex exponential modulated filter bank disclosed in the patent document (PCT / SE02 / 00626 “Reduction of aliasing using a complex exponential modulated filter bank”) can be expressed by the following complex analysis.

(9)

(9)

로부터 합성은 다음과 같이 정의된다.Where g _a is a fixed real analysis gain factor. Complex Subband Signal

Synthesis is defined as follows.

(10)

10

이고 코사인 변조 뱅크와 사인 변조 뱅크가 완전 재현성을 가진다고 가정하면, 수학식 8과 수학식 9로부터 다음이 얻어진다.Where g _s is a fixed real synthesis gain factor. Complex subband signal is unmodified

And assuming that the cosine modulation bank and the sinusoidal modulation bank have full reproducibility, the following is obtained from equations (8) and (9).

(11)

(11)

Here, complete reproducibility is obtained under the following conditions.

(12)

(12)

이다. In a particularly preferred embodiment, the fixed gain that implements energy conservation of the complex subband representation is

to be.

In the case of complex, since the deformation of the complex subband signals of Equations 9 and 10 can be omitted, the reproducibility is different from the specific modulation represented by Equation 4 as a fixed phase factor for each subband. It may be acceptable if it is not changed. The complex exponential modulated filter bank is oversampled twice. With proper window design, virtual aliasing-cancellation envelope adjustment as shown in patent document PCT / SE02 / 00626 “Reduction of Aliasing with Multiple Exponential Modulation Filter Banks” can be implemented. Such a design can often be easily obtained by abandoning the implementation of a strict full reproducibility framework as described above using an approximate full reproducibility implementation.

Multi-band Filter

만이 가능하다고 가정하면, 대응하는 사인 변조 뱅크 분석

은 코사인 뱅크 합성 단계와 사인 뱅크 분석 단계의 결합에 의해 얻어진다. 이에 따라, 다음이 얻어진다.Cosine Modulation Bank Analysis of Equation 7

Assuming only possible, analyze the corresponding sine modulation bank

Is obtained by combining the cosine bank synthesis step and the sine bank analysis step. Thus, the following is obtained.

(13)

(13)

Here, the change of the time variable of the dot product is defined as follows.

(14)

(14)

Thus, the summation for variable k in equation 13 corresponds to filtering, and the overall structure is understood as a version of multiband filtering with countless bands shown in FIG. When described again as a complex waveform, the following is obtained.

(15)

(15)

과 같이 대체하면, 수학식 15의 제1항은 다음과 같이 전개될 수 있다.

Substituted as follows, the first term of Equation 15 can be developed as follows.

(16)

(16)

를 가진 수학식 16의 적분은 소멸되고, 따라서 다음이 얻어진다. Symmetrical windows

The integral of Equation 16 with is eliminated, thus the following is obtained.

(17)

(17)

here,

(18)

(18)

This equation is the right function for μ and λ. By designing the appropriate window, one can assume that h _μ disappears at | μ |> 1. In the case of discrete time, the integral of Equation 18 can be replaced by the summation to the integer ν 'with the variable t = ( v + θ ) / L. Where L is the number of subbands and θ is an offset value of 0 or 1/2. In Equation 18, the discrete time corresponding part is a periodic function having a period 2L at θ = 0 and an aperiodic function having a period 2L at θ = 1/2. Substituting n = m + r in (15), the following is obtained.

(19)

(19)

Referring to the component 402m of FIG. 4, in the above calculation, L = K is substituted, and f _{m, r} (λ) = <c _{m + r} , s _{m , λ} > is an impulse of the filter F _{m , r} . Can be used as a response. When h _μ is extinguished except in the case of μ = 2Kκ + σ, where κ is an integer and σ∈ {-1,0,1}, the second term in equation (19) is m = 0 and m = (K− This boundary is important because it provides a kind of key to the near invertability of the multi-band filter 401. Unlike the simple variant of Equation 19, Only two standard filters h ₀ and h ₁ should be considered, and if we look at equation 19, only the sample of the function of the standard filter h ₀ will contribute. It will be clear to those skilled in the art that the similarity of the filters f _{m + 1, -1} and f _{m -1,1} makes it possible to implement a multiband filter very efficiently in a multiphase form. The application will be described in more detail.

In practical design, it is effective to abandon the discrete product of Equation 18 for the design of a standard filter. Instead, for a given integer N, the filters f _{m , r} are designed with the best approximation:

(20)

20

This provides a second direct path to the sine modulation bank analysis.

(21)

(21)

Here, the asterisk means convolution. By developing the sine synthesis operation of Equation 8, substituting it into Equation 20, and combining the cosine terms, the following is obtained.

(22)

(22)

Thus, synthesizer multiband filter 306 also imposes an impulse response. It has the structure of a multi-band filter 401 with a filter replaced with G _{m , r} with G _{m , r} (λ) = f _{m + r, -r} (-λ). The same result is obtained by swapping the roles of cosine modulation and sine modulation in the above-described derivation process.

The overall computational complexity of the multiband filter is proportional to the number of N * K operations per subband sample period, i.e., N * K / L operations per digital audio sample. If K << L, a significant amount of computation can be reduced compared to the additional sine modulation required for a full complex modulation filter bank.

Compared with the application of pure real or pure imaginary modulation filter banks, additional delays of N subband samples are introduced by the multiband filter in the analysis and synthesis phases. This may be compensated for by delaying all subband samples that did not pass the multiband filter as delays of N subband samples in delayers 202, 203, 304, and 305. Deformation processing 103 is described in E. Schuijers, J. Breebart, H. Purnhagen, J. Engdegard: “Low Complexity Parametric Stereo Coding”, Proc. 116th AES convention, 2004, paper 6073. In the case of subband filtering, sub-subband filtering can be combined with the multiband filter 204 to approximate the combined impulse response to reduce the overall delay.

In the case where the selected K complex subbands are the first K subbands of the total L subbands, the multiband filter takes the effect of synthesizing the filter bank with K subbands K / L times the raw sampling frequency. Emulate into domain and perform analysis with filter bank with K subbands. This bypass operation has the problem that the multiband filter is longer than can be obtained by the design method according to the present invention. In applications where the number of analytical audio channels is much smaller than the number of synthesis channels, partial complex analysis 101 is performed by a pure imaginary modulated filter bank analysis with L subbands and an imaginary part with the last (LK) subbands. By eliminating, the computational complexity is increased, but the analysis delay of the multiband filter can be avoided. However, in order to combine the synthesis of FIG. 3 to achieve an approximate full reproducibility in the case of unaltered subbands, an analysis of the boundary subband with exponent (K-1) is followed by a special direct It is necessary to replace with type filter. Thus, such a filter can be obtained by examining the partial complex synthesis of FIG. 3 where the boundary subband with exponent K-1 includes nonzero samples and the other subbands are zero. Although more useful in terms of complexity reduction, the synthesis delay of multiple filter bands is L subbands that are sent back to a separate synthesis operation consisting of upsampling with the input subband having an exponent (K-1) factor L. This can be avoided likewise by performing partial complex synthesis 104 by pure imaginary modulation filter bank synthesis with and thus performing special direct filtering. The result of the complex bank synthesis for the (L-1) bands and the synthesis of the separate bands is added to the time domain.

The system according to the present invention includes frequency selective spatialization, frequency selective panning, spectral envelope adjustment, or equalization of an audio signal using downsampled real value subband filter banks. Accordingly, aliasing in the selected frequency range can be suppressed by converting the corresponding subset of subband signals into a complex subband signal. Under the assumption that aliasing outside the selected frequency range is hardly detected or otherwise mitigated, the present invention has the effect of greatly reducing the amount of calculation compared to the case of using a complex filter bank.

The present invention discloses a method of obtaining a complex representation for a signal of a certain frequency range with computational complexity slightly greater than that of a real filter bank. An effective multiband filter is applied to the selected subbands of real filter bank analysis to produce an imaginary part of the subband signal. The result is a partial complex modulation filter bank analysis. The complex subband signal will have the same advantages as the corresponding subband from the complex exponential modulated filter bank in terms of minimal aliasing and energy stability assessment resulting from linear time invariant deformations such as envelope adjustment and filtering. Before performing real synthesis, another multiband filter converts the complex subband samples back to real subband samples. The overall performance of the reproduction and signal processing is comparable to the performance of the complex filter bank in the complexed frequency range, and also to the performance of the real filter bank in the remaining frequency range. Seamless transitions between two frequency ranges are inherently implemented by the special boundary band processing disclosed herein.

In the frameworks of the transducers to manipulators 102 and 103, it is necessary to mention temporal interpolation gains or time-varying applications of spatial parameters through the matrix (eg MEPG surround to parametric stereo). In the case of time-invariant deformation or manipulation, application to envelope regulation or equalization is important as a feature to prevent the introduction of aliasing. Thus, in time-invariant cases, the definition of the introduction of aliasing is of primary interest.

However, introducing a time-varying configuration into the framework of the manipulators and transducers 102 and 103 shown in FIG. 1 makes it difficult to define features for preventing the introduction of aliasing. Indeed, the long main parts of the signal will be handled locally in an invariant way in the framework of MEPG surround. As an additional step, nonlinear manipulation is also considered in frameworks with advanced transition methods such as high performance SBR, which is very important. This improved transition method includes time varying manipulation and / or nonlinear manipulation, but time invariant deformation and manipulation should be considered in the first step.

In summary, in the frameworks of the transformers to manipulators 102 and 103, any operation can be reliably made as long as the time-frequency resolution of the resulting (partially complex) filter bank is as long as possible. The advantages of the operation 103 of the corresponding complex bank all appear in the complex part of the partial complex filter bank.

Embodiments of the invention shown in Figures 1 to 3 include the following features.

A method of transforming a discrete time audio signal,

Filtering the discrete time audio signal by a cosine modulation analysis filter bank;

Generating a complex subband sample for the subband subset by multiband filtering;

Transforming the real subband sample and the complex subband sample;

Converting the complex sample into a real sample by multiband filtering; And

Filtering the real subband samples through a cosine modulated synthesis filter bank to obtain a modified discrete time audio signal.

The implementation of a low power version of the spatial audio tool is briefly described below. The low-power spatial audio tool processes real-valued subband domain signals above the Kth QMF subband (QMF = quadrature mirror-filter). Where K is a positive integer. The constant K is chosen according to the specific requirements and requirements of the intended configuration. In other words, it is given by the details of the intended configuration, such as bitstream information. Real-valued QMF filter banks are used with the real-complex converter, a feature of the present invention, to obtain a partial complex subband domain representation. The low power spatial audio tool may also include additional modules to reduce aliasing introduced in the real value processing.

Following this simple introduction, the low power spatial audio coding system comprises a time / frequency converter according to FIG. 10. The time / frequency converter of the spatial audio coding system includes the hybrid QMF analysis bank shown in FIG. The hybrid QMF analysis bank is connected to the real-complex converter 520 that is a feature of the present invention through an optional switch 510 for processing the real-time QMF analysis bank 500. Real-complex converter 520 is connected to one or more nequist analysis banks 530.

를 제공받아, 그 출력으로서 실수치 QMF 신호

를 실수-복소 컨버터(520)에 제공한다. 실수-복소 컨버터(520)는 QMF 신호를 부분 복소 샘플

^n,m로 전환하고, 이들은 니퀴스트 분석 뱅크(530)에 제공되며, 이들 뱅크는 하이브리드형 부대역 도메인 신호 x ^{n ,m}를 생성한다.The real value QMF analysis bank 500 is a time domain input signal as its input.

Is supplied as a real valued QMF signal

To the real-complex converter 520. Real-complex converter 520 partially complex samples the QMF signal.

^{n, m} , which are provided to the Nyquist analysis bank 530, which banks the hybrid subband domain signal Create x ^{n , m}

가 설정되고, (중간) 실수치(QMF) 부대역 도메인 샘플

가 예로서 낮은-복잡도 HE-AAC 디코더로부터 취해질 수 있다. 좀더 구체적으로 말하면, 문헌(ISO/IEC 14496-3:2001/AND1: 2003)에 개시된 바와 같이, 부대역 도메인 샘플이 HE-AAC QMF 합성전에 취해지는 경우가 이에 해당한다. 이들 QMF 입력 신호

가 본 발명의 특징인 실수-복소 컨버터(520)에 공급되도록 하기 위해, 선택사항인 스위치(510)가 도 10에 도시된 시간/주파수 변환기에 통합되어 스위칭 동작을 한다.Apart from the usual mode of operation of the time / frequency converter, the spatial audio decoder has time domain samples.

Is set, and (middle) real (QMF) subband domain samples

Can be taken as an example from a low-complexity HE-AAC decoder. More specifically, this is the case when subband domain samples are taken prior to HE-AAC QMF synthesis, as disclosed in ISO / IEC 14496-3: 2001 / AND1: 2003. These QMF input signals

An optional switch 510 is integrated into the time / frequency converter shown in FIG. 10 to perform a switching operation so that is supplied to the real-complex converter 520 that is a feature of the present invention.

^n,m로 변환된다. 또한, 가능하다면 추가적인 선택사항으로, 도 10에 도시되지 아니한 잔여 디코딩 모듈이 부대역 도메인 샘플

를 QMF 잔여 입력 신호로서 제공할 수 있다. 하이브리드형 부대역 도메인 신호 x^{n ,m}를 형성하는 하이브리드형 도메인으로 변환되기 전에, 이들 QMF 잔여 신호는 또한 선택사항인 지연기(540)를 통하여 니퀴스트 분석 뱅크(530)로 통과되고, 한편 이들 QMF 잔여 입력 신호는 또한 실수-복소 컨버터(520)의 기인하는 지연을 보상하기 위해 지연된 형식으로 통과될 필요가 있다.Real-valued QMF samples provided in the form of a QMF input signal or through the real QMF analysis bank 500 are partially complex samples by the real-complex converter 520 described below with reference to FIG.

is converted to ^{n, m} . Also, if possible, an additional option is to add a subband domain sample to the residual decoding module, not shown in FIG.

May be provided as the QMF residual input signal. Before being converted to the hybrid domain forming the hybrid subband domain signal x ^{n , m} , these QMF residual signals are also passed through the optional retarder 540 to the Nyquist analysis bank 530, while these The QMF residual input signal also needs to be passed in a delayed form to compensate for the delay caused by the real-complex converter 520.

^n,m로 변환된다. 부분 복소 QMF 부대역 도메인 샘플은, 부분 복소 QMF 부대역 도메인 샘플을 실수치 내지 실수 QMF 샘플

로 변환하는 본 발명의 특징인 복소-실수 컨버터(560)에 제공된다. 본 발명의 특징인 복소-실수 컨버터(560)는 도 14를 참조하여 더욱 상세히 설명될 것이다. 이들 실수치 QMF 샘플은 실수치 QMF 합성 뱅크(570)에 제공되어, 시간 도메인 샘플 내지 시간 도메인 출력 신호

의 형태로 시간 도메인으로 다시 변환된다.11 illustrates a hybrid QMF synthesis bank for performing frequency / time conversion to time / frequency conversion in a spatial audio coding system. The hybrid QMF synthesis bank includes one or more nequist synthesis banks 550 to which a hybrid subband domain signal y ^{n , m} is provided at the input. More specifically, on the Nyquist synthesis side, the hybrid subband domain samples y ^{n , m} are partially complex QMF subband domain samples by the Nyquist synthesis bank 550.

is converted to ^{n, m} . The partial complex QMF subband domain sample is a real or real QMF sample

To complex-real converter 560, which is a feature of the present invention that converts to &lt; RTI ID = 0.0 &gt; Complex-real converter 560, a feature of the present invention, will be described in more detail with reference to FIG. These real QMF samples are provided to a real QMF synthesis bank 570 to provide time domain samples to time domain output signals.

Is converted back to the time domain in the form of.

를 64개의 부대역 신호로 분할하는 데에 이용된다. 필터 뱅크 내지 실수치 QMF 뱅크(500)로부터의 출력은 실수치이며 부대역 샘플의 형식으로 임계적으로 샘플링된 신호이다.Hereinafter, the filter bank, more specifically, the real value QMF analysis bank 500 and the real value QMF synthesis bank 570 will be described in more detail. As an example, a real QMF filter bank is used for a low power MPEG surround system. In this case, as explained below, the analysis filter bank 500 uses 64 channels. The synthesis filter bank 570 also has 64 channels and is equivalent to the filter banks used in low complexity HE-AAC systems, which are disclosed in section 4.6.18.8.2.3 of ISO / IEC 14496-3. . Hereinafter, description will be made based on 64 channels (integer L = 64), but the present invention and embodiment are not limited to using only 64 channels and a specific appropriate number of real and complex subband signals. In principle, any number of channels or real to complex subband signals may be used in embodiments of the present invention. However, if other numbers of channels are used, the appropriate parameters of this embodiment will also need to be changed. The real QMF analysis bank 500 shown in FIG. 10 is a time domain signal from the core decoder.

Is used to divide the signal into 64 subband signals. The output from the filter bank or real value QMF bank 500 is a real value and is a signal that is critically sampled in the form of subband samples.

12 shows a flow diagram of the operations performed by real-value analysis QMF bank 500 in the form of C / C ++ pseudocode. In other words, the method performed by the real valued QMF analysis bank 500 is shown in FIG. 12. The filtering involves the following steps, where the array x contains 640 time domain input samples indicated by exponents in the range 0 to 639. In FIG. 12, the exponents of the arrays to the vector are indicated in square brackets. If the exponent for the array x of time domain input samples is high, it means older samples.

12 shows the method performed by the real valued QMF analysis bank 500 for QMF subband sample 1. After initiating the method of step S100, the samples in the array x are shifted by 64 positions in step S110. The oldest 64 samples with exponents ranging from 575 to 639 (n = 575, ..., 639) are removed. Next, in step S120 64 new samples are stored at the positions indicated by indices 0 to 63 in the arrangement x.

In step S130, the samples of the array x are multiplied by the set of coefficients of the window to window function c. Window c is also implemented as an array c with 640 components indicated by exponents in the range n = 0, ..., 639. This multiplication is performed by introducing a new intermediate arrangement z with 640 components in step S130 as follows.

z (n) = x (n) * c (n) (where n = 0, ..., 639) (23)

Here, the window coefficients c [0], ..., c [639] are shown in Table 4.A.87 of the document (ISO / IEC 14496-3).

In a next step S140, the samples represented by the intermediate array z are summed as follows.

(24)

(24)

This creates a new intermediate array u with 128 components. Equation 24 is also shown in the flowchart of FIG. 12 as a mnemonic code representing equation (24).

In the next step S150, 64 new subband samples are calculated by the matrix operation M * u using the matrix M defined as follows.

(25)

(25)

This is performed before performing the filtering method in step S160.

와 동일하다This produces 64 subband samples, each loop representing the output from the filter bank subband, in each loop of the method shown in the flowchart of FIG. As described above in the flowchart of FIG. 12, X _real [m] [l] corresponds to subband sample I of QMF subband m. Where m, l, and n are all integers. Therefore, the output X _real [m] [n] is the real-valued subband sample

Is the same as

FIG. 12 shows a flowchart of a real value analysis QMF bank 500, and FIG. 13 shows in more detail the real-complex converter 520 that is a feature of the present invention shown in FIG. The real-complex converter 520 shown in FIG. 13 has 64 real-valued units forming two distinct subsets of a subset of the K real subbands and a subset of the real subbands of 64-K. Receive a reverse signal. Here, K is a positive integer in the range of 1 to 64. The K real number subband signals to a subset of subbands form a plurality of real value subband signals, and the second subset of (64-K) real subbands outputs a number of other real number subband signals. Form.

A subset of the K real valued subband signals is provided to both the multiband filter 600 and the optional first delayer 610. The multi-band filter 600 provides a set of K real valued intermediate subband signals at the output, which multiplier 620 multiplies each of the real valued intermediate subband signals by a negative imaginary unit (-i). Is provided. The output of multiplier 620 is provided to adder 630 which receives K real number subband signals from delayer 610 in a delayed form. The output of adder 630 is also provided to fixed gain regulator 640. The fixed gain regulator 640 adjusts the level of each subband signal provided to the input side by multiplying the real value constant by the corresponding subband signal. It should be noted that the fixed gain regulator 640 is an optional component and is not an essential component of the real-complex converter 520 that is a feature of the present invention. If a fixed gain regulator 640 is provided, at the output of that fixed gain regulator, i.e., at the output of the adder 630, the real-complex converter 520 may have K complex subband signals to K complex subbands. To provide.

The adder 630, together with the multiplier 620, constitutes a calculator 650, which provides a complex subband signal, whose gain is adjusted by an optional fixed gain adjuster 640. More specifically, the calculation unit 650 is a real part of the complex subband signal as the real part of the complex subband signal and the imaginary part of the complex subband signal output by the calculation unit 650 and the multiband filter 600. Combine the intermediate signals output by.

It should be noted here that the first delayer 610 is also an optional component, which is provided by the calculator 650 to the intermediate signal output by the multi-band filter 600 and the real-complex converter 520. Before combining the real valued subband signals, it is ensured to accurately account for possible time delays due to the multiband filter 600.

As an optional component, the real-to-complex converter 520 also includes a second delayer 660, which has a potential time delay due to the multi-band filter 600 to a large number of real values. Ensure that (64-K) real values of the subband signal do not appear in the subband signal. To this end, a second delay 660 is connected between the 64 real-valued subband signals passing through the real-complex converter 520 (64-K) in an unaltered form. Here, the real-complex converter 520 does not necessarily have to contain a real-valued subband signal transmitted in an unaltered or merely delayed form, so that any real-valued subband signal is not included in the real-complex converter 520. It is important to note that may not pass in the form described above. Here, the integer K can also be assumed to be K = 64.

The real QMF subband signals are converted into partial complex QMF subband signals by the real-complex converter 520 shown in FIG. The K real number subband signals of the first group are filtered by the multiband filter 600, the negative imaginary unit (-i) is multiplied by the multiplier 620, and the K delayed by the adder 630. In addition to the real value subband scene, K complex subband signals are obtained. As discussed above, a delay 610 may be provided as an option to delay the K real valued subband signals before being processed by the adder 630. The K complex subband signals output by the adder 630 or the calculation unit 65 are adjusted by the fixed real gain regulator 640, and the real-complex converter 320 and the real-complex converter thereof. It is output as K complex subbands of a partial complex analysis filter bank comprising a.

는 계산부(650)에 의해 출력된 복소치 부대역 신호의 허수부인, 다중 대역 필터(600)의 출력을 나타낸다.A second group containing (64-K) real subband signals may be delayed by the second delayer 660 if provided with an optional second delayer 660. Optional delay 610 and delay 660 serve to compensate for the potential delay introduced by multiband filter 600. Typically, the delay length is related to the order of the multiband filter set included in the multiband filter 600. The standard delay length is half of the order of the standard public band filter. This means that the delay imparted by the optional two delayers 610 and 660 according to the invention described in more detail below corresponds to five subband samples. As described above, particularly as in the description of the multiband filter shown in FIG. 4, the multiband filter performs the following calculation on the K first QMF subband signals. here,

Denotes the output of the multi-band filter 600, which is an imaginary part of the complex subband signal output by the calculator 650.

(26)

(26)

는 다중 대역 필터에 입력되는 실수치 부대역 신호를 나타낸다. 또한, QMF 부대역의 합산의 상한과 하한은 다음과 같이 정의된다.Where f _{m , r} [ν] term represents a filter or filter function,

Denotes a real value subband signal input to the multi-band filter. In addition, the upper and lower limits of the sum of the QMF subbands are defined as follows.

(27)

(27)

(28)

(28)

The filters f _{m , r} [ν] are derived from two standard multiband filters 600, which are generally determined by two standard multiband filters a ^ν [n], where ν = 0,1. do. More specifically, those filters or filter functions perform the following relationship.

(29)

(29)

Here, the standard multi-band filter coefficients a ⁰ [ν] have a relationship given in Table 1 below.

0.003 ≤ a ⁰ [0] ≤ 0.004 A ⁰ [1] | ≤0.001 -0.072 ≤ a ⁰ [2] ≤-0.071 A ⁰ [3] | ≤0.001 0.567≤a ⁰ [4] ≤0.568 A ⁰ [5] | ≤0.001 0.567≤a ⁰ [6] ≤0.568 A ⁰ [7] ｜ ≤0.001 -0.072≤a ⁰ [8] ≤-0.071 A ⁰ [9] ｜ ≤0.001 0.003≤a ⁰ [10] ≤0.004

In addition, the standard multi-band filter coefficients a ¹ [ν] have a relationship given in Table 2 below.

0.0008 ≤ a ¹ [0] ≤ 0.0009 0.0096 ≤ a ¹ [1] ≤ 0.0097 0.0467 ≤ a ¹ [2] ≤ 0.0468 0.1208 ≤ a ¹ [3] ≤ 0.1209 0.2025 ≤ a ¹ [4] ≤ 0.2026 0.2388 ≤ a ¹ [5] ≤ 0.2389 0.2025 ≤ a ¹ [6] ≤ 0.2026 0.1208 ≤ a ¹ [7] ≤ 0.1209 0.0467 ≤ a ¹ [8] ≤ 0.0468 0.0096 ≤ a ¹ [9] ≤ 0.0097 0.0008 ≤ a ¹ [10] ≤ 0.0009

In other words, the filter f _{m , r} [ν] is derived from Equation 29 from the standard filters given in Tables 1 and 2.

은 지연된 실수치 QMF 부대역 샘플

와 결합되어, 도 13에 도시된 부분 복소 QMF 부대역 샘플

를 형성한다.. 좀더 구체적으로 말하면, 출력

는 다음과 같은 관계를 가진다.The output of the multi-band filter 600 by the calculation unit 650

Is a delayed real value QMF subband sample

Combined with, the partial complex QMF subband sample shown in FIG.

To be more specific, the output

Has the following relationship:

(30)

(30)

의 위첨자(n-5)는, 두 지연기(610 및 660)의 영향을 나타낸다. 상술한 바와 같이, 표준적인 지연 길이는 표 1 및 2로 주어진 표준적인 다중 대역 필터 계수 a^ν[n]의 차수의 절반에 해당한다. 또한, 이는 5개의 부대역 샘플에 해당한다.Where the real QMF subband sample

Superscript n-5 denotes the effect of the two retarders 610 and 660. As mentioned above, the standard delay length corresponds to half of the order of the standard multi-band filter coefficients a ^v [n] given in Tables 1 and 2. This also corresponds to five subband samples.

In another embodiment of the present invention, the standard multi-band filter or the standard multi-band filter coefficients a ^v [n] (where v = 0,1) have a relationship given in Tables 3 and 4 below.

0.00375672984183 ≤ a ⁰ [0] ≤ 0.00375672984185 A ⁰ [1] | ≤0.00000000000010 -0.07159908629243 ≤ a ⁰ [2] ≤-0.07159908629241 A ⁰ [3] | ≤0.00000000000010 0.56743883685216 ≤ a ⁰ [4] ≤ 0.56743883685218 A ⁰ [5] | ≤0.00000000000010 0.56743883685216 ≤ a ⁰ [6] ≤ 0.56743883685218 A ⁰ [7] | ≤0.00000000000010 -0.07159908629243 ≤ a ⁰ [8] ≤-0.07159908629241 A ⁰ [9] ｜ ≤0.00000000000010 0.00375672984183 ≤ a ⁰ [10] ≤ 0.00375672984185

0.00087709635502 ≤ a ¹ [0] ≤ 0.00087709635504 0.00968961250933 ≤ a ¹ [1] ≤ 0.00968961250935 0.04670597747405 ≤ a ¹ [2] ≤ 0.04670597747407 0.12080166385304 ≤ a ¹ [3] ≤ 0.12080166385306 0.20257613284429 ≤ a ¹ [4] ≤ 0.20257613284431 0.23887175675671 ≤ a ¹ [5] ≤ 0.23887175675673 0.20257613284429 ≤ a ¹ [6] ≤ 0.20257613284431 0.12080166385304 ≤ a ¹ [7] ≤ 0.12080166385306 0.04670597747405 ≤ a ¹ [8] ≤ 0.04670597747407 0.00968961250933 ≤ a ¹ [9] ≤ 0.00968961250935 0.00087709635502 ≤ a ¹ [10] ≤ 0.00087709635504

In another embodiment of the invention, the standard multi-band filter coefficients a ^v [n], where v = 0,1, comprise the values given in Table 5 below.

n a ⁰ [n] a ¹ [n] 0 0.00375672984184 0.00087709635503 One 0 0.00968961250934 2 -0.07159908629242 0.04670597747406 3 0 0.12080166385305 4 0.56743883685217 0.20257613284430 5 0 0.23887175675672 6 0.56743883685217 0.20257613284430 7 0 0.12080166385305 8 -0.07159908629242 0.04670597747406 9 0 0.00968961250934 10 0.00375672984184 0.00087709635503

As outlined in the mathematical background, in particular with respect to equations (18) to (20), the structure of the coefficients a ^v [n] results in symmetry. More specifically, the coefficient a ^ν [n] given in Table 5 above has the following symmetrical relationship.

a ^ν [10-n] = a ^ν [n] (30a)

Where ν = 0, 1, n = 0, ..., 10,

a ⁰ [2n + 1] = 0 (30b)

Where n = 0, ..., 4

Referring to FIG. 11, in a previous step of the real valued QMF synthesis section 570, the partial complex subband QMF signal is converted by the complex-real number converter 560 into a real valued QMF signal, which is shown in more detail in FIG. It is.

The complex-to-real number converter 560 shown in FIG. 14 includes 64 subband signals including K complex subband signals and (64-K) real subband signals. Multiple K complex subband signals, or other K complex subbands, are provided to the fixed gain adjuster 700, which is an optional component of the majority-real converter 560. As described above, K represents a positive integer in the range of 1 to 64. In addition, the present invention is not limited to 64 subband signals, and may process 64 or more subband signals. In this case, the parameters of the present embodiment disclosed next may also be appropriately modified.

The fixed gain adjuster 700 is connected to the separator 710 to the extractor 710, and as described above, the extractor 710 receives the output of the fixed gain adjuster 700 as its real part extractor 720. And an imaginary part extractor 730. However, unless the optional fixed gain regulator 700 is implemented, the separators 710 to extractor 710 directly receive K complex subband signals. Real part extractor 720 is connected to an optional first delayer 740, and an imaginary part extractor 730 is connected to multiband filter 750. First delay 740 and multiband filter 750 is coupled to a calculator 760 that provides K real subband signals as an output of the complex-real converter 560 that is a feature of the present invention.

In addition, the complex-to-real converter 560 is provided with (64-K) real subband signals, described as real subbands in FIG. 14, which are also provided to the optional second delay unit 770. do. At the output of the complex-real converter 560, (64-K) real subband signals are provided in delayed form. However, if the second delayer 770 is not provided, the (64-K) real valued subband signals are passed without modification.

즉, K개의 복소치 부대역 신호의 복소부는 고정 이득 조절기(700)에 의해 그 이득이 조절된다. 고정 이득 조절기(700)는 입력되는 모든 복소치 부대역 신호에 실수치 인수(factor), 즉, 1/√2를 승산한다. 다음에, 분리기(710)는 실수부 추출기(72)과 허수부 추출기(730)을 이용하여 다음의 수학식에 따라 그 이득 조절된 신호를 실수부 신호

와 허수부 신호

로 분리한다.In the embodiment shown in Figure 14, the partial complex QMF subband signal

That is, the complex portion of the K complex subband signals is adjusted by the fixed gain adjuster 700. The fixed gain adjuster 700 multiplies all the complex subband signals that are input by a real value factor, that is, 1 / √2. Next, the separator 710 uses the real part extractor 72 and the imaginary part extractor 730 to convert the gain adjusted signal according to the following equation.

And imaginary signal

To separate.

(31)

(31)

의 앞에 기재된 인수인 1/√2는 고정 이득 조절기(700)에 의해 제공된다.In the embodiment shown in FIG. 13, a complex subband signal

The 1 / √2 argument described before is provided by the fixed gain regulator 700.

에 대해 다음과 같은 수학적 동작을 수행한다.The multi-band filter 750 is an imaginary part signal that is a real value signal

Perform the following mathematical operations on.

(32)

(32)

의 일 세트를 제공한다. 수학식 32에서, QMF 부대역 합산의 상한 p(m)과 하한 q(m)은 각각 상술한 수학식 27과 28에 의해 정의된다. 또한, 필터 내지 필터 함수 g_{m ,r}[ν]는 표 2, 표 3, 및 표 4, 또는 표 5에 도시된 표준적인 필터 내지 필터 계수로부터 다음과 같은 관계식을 통하여 도출된다.Multi-band filter 750 is a K real number intermediate subband signal

Provide a set of days. In Equation 32, the upper limit p (m) and the lower limit q (m) of the QMF subband summation are defined by Equations 27 and 28, respectively. Further, the filter to filter function g _{m , r} [ν] is derived from the standard filter to filter coefficients shown in Tables 2, 3, and 4, or 5 through the following relationship.

(33)

(33)

를 얻기 위하여, 계산부(760)는 다중 대역 필터(750)로부터 출력되는 중간 부대역 신호와 분리기(710)로부터 지연된 형식으로 출력되는 실수부 신호를 합산한다.QMF signal for K complex subband signals processed by separator 710 or extractor 710 and multiband filter 750

In order to obtain, the calculation unit 760 sums the intermediate subband signal output from the multi-band filter 750 and the real part signal output in a delayed form from the separator 710.

가 다음과 같은 연산을 수행하여 얻어진다.The remaining (64-K) real subband signals are passed in a delayed fashion under the influence of the second delay unit 770. In summary, the QMF signal to be supplied to the real valued QMF synthesis bank 570 shown in FIG.

Is obtained by performing the following operation.

(34)

(34)

및 실수치 부대역 신호

의 위첨자는 제1 지연기(740) 및 제2 지연기(670)에 기인하는 것이고, 이들의 지연 길이는 전형적으로 표 1 내지 5에 주어진 표준적 다중 대역 필터(a^ν[n])의 차수의 절반에 해당한다. 상술한 바와 같이, 이는 5개의부대역 샘플에 해당한다.As described above with reference to Equation 30, the real part signal

And real-valued subband signals

The superscript of is due to the first delayer 740 and the second delayer 670, the delay length of which is typically the order of the standard multiband filter (a ^v [n]) given in Tables 1-5. Corresponds to half. As mentioned above, this corresponds to five subband samples.

As described above with reference to FIG. 13, the present invention is not limited to only 64 subband signals to K complex subband signals. Indeed, if the number K of complex subband signals is equal to the number of all subband signals (K = 64), then the second delayer 770 may also be omitted, like the second delayer 660 of FIG. have. Thus, the total number of subband signals (integer L = 64) is not limited and is not required. With proper adjustment of the parameters of the components shown in FIG. 14, in principle, any number L of subband signals can be used as input to the complex-real converter 560.

The invention is not limited to multiband filters 204, 306, 401, 600, and 750, which operate to symmetrically distribute the subband signals associated with the exponent m over the subbands. In other words, the present invention combines a subband signal having an exponent distributed symmetrically with respect to the exponent of an intermediate subband signal output by a multi-band filter or other signals, for example, an exponent m and an integer m '. It is not limited to multiband filters, starting with subbands with exponents m, (m + m '), and (m-m') from subbands with. Apart from the obvious constraints on subband signals with such exponents, where the exponent is small or large so that the subband signal cannot be selected symmetrically, the multiband filter has a respective intermediate subband output from the multiband filter. It can be designed to use individual combinations of subband signals for inverse signals. In other words, the number of subband signals processed to obtain the intermediate subband signal may not be three. For example, if different filters with different filter coefficients are selected, as described above, it is preferable to use more than three subband signals in total. In addition, the multiband filter may be designed to provide or output an intermediate subband signal having an exponent that does not correspond to the exponent of the subband signal provided to the multiband filter. In other words, if a multiband filter outputs an intermediate subband signal with an exponent m, then a subband signal with the same exponent is not necessarily required as any subband signal provided to that multiband filter.

In addition, a system including any or all of converters 520 and 560 may include additional aliasing detectors and / or aliasing equalizers to aliasing equalization means.

In a preferred embodiment, the device (210; 520) operates to assign an exponent m to each of the real subband signals according to a center frequency associated with the real subband signal, thereby increasing the real value with an increasing exponent m. A subband signal is arranged according to a center frequency associated with the real subband signal, wherein the plurality of real subband signals comprise K real subband signals, where K is a positive integer and m is It is an integer in the range from 0 to (K-1).

In a preferred embodiment, the apparatus 310 (560) according to the present invention comprises a plurality of complex subband signals each comprising a first complex subband signal and a second complex subband signal to obtain a real value subband signal. A device (310; 560) for processing a first imaginary part from the first complex subband signal, a second imaginary part from the second complex subband signal, and the plurality of complex subbands. An extractor (309; 710) for extracting a real part from the first, second, or third complex subband signal of the signal; Filter the first imaginary part to obtain a first filtered imaginary part signal, filter the second imaginary part to obtain a second filtered imaginary part signal, and filter the first imaginary part to obtain an intermediate subband signal A multi-band filter (306; 750) for combining the signal with the second filtered imaginary signal to provide the intermediate subband signal; And a calculator 307 (760) for combining the real part and the imaginary part to provide the real value subband signal.

을 조절하는 이득 조절기(301; 700)을 포함한다.In a preferred embodiment, the device (310; 560) is a value of a complex subband signal of the plurality of complex subband signals.

It includes a gain regulator (301; 700) for adjusting the.

In a preferred embodiment, the device (310; 560) further delays the real part signal and further delays (305; 740) for passing the real part signal to the multi-band filter (306; 750) in a delayed form. It is a device that includes.

In a preferred embodiment, the extractors 309 and 710 operate to extract a first real part from the first complex subband signal and to extract a second real part from the second complex subband signal.

In a preferred embodiment, the multi-band filter 306 (750) is a low-pass filter, high pass filter, or bandpass filter for filtering the first imaginary signal and the second imaginary signal It adopts and operates.

In a preferred embodiment, the apparatus 310; 560 is operative to assign an exponent m to each complex subband signal of the plurality of complex subband signals according to a center frequency associated with the complex subband signal, The complex subband signal with increasing exponent m is arranged according to a center frequency associated with the complex subband signal, wherein the plurality of complex subband signals comprise K complex subband signals, and the K Is a positive integer and m is an integer in the range from 0 to (K-1).

를 가진 각 복소치 부대역 신호에 대하여

값 을 가진 실수치 실수부 신호 및 값

을 가진 실수치 허수부 신호를 공급하도록 동작하고, 상기

,

, 및

는 다음의 수학식에 기초하는 관계를 수행한다.In a preferred embodiment, the extractors 309; 710 have a value in the plurality of complex subband signals with an exponent m in the range of 0 to (K-1).

For each complex subband signal with

Real value with value Real part signal and value

Operate to supply a real imaginary imaginary signal having

,

, And

Performs a relationship based on the following equation.

In a preferred embodiment, the extractors 309 and 710 associate each imaginary part signal and / or real part signal with the exponent m of the complex subband signal separated into the imaginary part signal and / or the real part signal. To work.

In a preferred embodiment, the multiband filter 306 (750) operates to associate the exponent m with the intermediate subband signal corresponding to the exponent m of the first imaginary part signal.

In a preferred embodiment, the multiband filter 306; 750 operates to use an imaginary part signal having an exponent (m + 1) or an exponent (m-1) as the second imaginary part signal, and the exponent m Is the exponent of the first imaginary part signal.

In a preferred embodiment, the multi-band filter 306; 750 is configured to correspond to the imaginary part of the third complex subband signal of the plurality of complex subband signals to obtain a third filtered imaginary part signal. And further filter the third imaginary part signal received from (309; 710), the first filtered imaginary part signal, the second filtered imaginary part signal, and the third to obtain the intermediate subband signal. Operate to combine the filtered imaginary part signal, wherein the second imaginary part signal is associated with the exponent (m-m ') and the third imaginary part signal is associated with the exponent (m + m'), or the first 2 an imaginary part signal is associated with the exponent (m + m ') and the third imaginary part signal is associated with the exponent (m-m'), m is an exponent of the first imaginary part signal, and m 'Is a positive integer.

In a preferred embodiment, the multiband filter 306 (750) operates to provide a real valued intermediate subband signal for each of the intermediate subband signals as the first intermediate subband signal having an exponent m.

을 가진 K개의 중간 실수치 부대역 신호를 제공하도록 동작하고, 상기 n 및 m은 정수이고, 0부터 (K-1)까지의 범위내의 상기 지수 m을 가진 상기 K개의 실수치 허수부 신호 각각에 대해 다음의 수학식이 정의되고, In a preferred embodiment, the multi-band filter 306 (750) is a value

Provide K intermediate real-valued subband signals, with n and m being integers, each of the K real-valued imaginary part signals having the exponent m in the range of 0 to (K-1). The following equation is defined for

Ν is an integer in the range from 0 to 10, wherein g ^{m , r} [n] is defined by the following equation,

A ⁰ [ν] and a ¹ [ν] are coefficients of the standard filter, and each of the coefficients a ⁰ [ν] and a ¹ [ν] of the standard filter follows the relationship of the following Tables 6 and 7.

0.003 ≤ a ⁰ [0] ≤ 0.004 A ⁰ [1] | ≤0.001 -0.072 ≤ a ⁰ [2] ≤-0.071 A ⁰ [3] | ≤0.001 0.567≤a ⁰ [4] ≤0.568 A ⁰ [5] | ≤0.001 0.567≤a ⁰ [6] ≤0.568 A ⁰ [7] ｜ ≤0.001 -0.072≤a ⁰ [8] ≤-0.071 A ⁰ [9] ｜ ≤0.001 0.003≤a ⁰ [10] ≤0.004

And

0.0008 ≤ a ¹ [0] ≤ 0.0009 0.0096 ≤ a ¹ [1] ≤ 0.0097 0.0467 ≤ a ¹ [2] ≤ 0.0468 0.1208 ≤ a ¹ [3] ≤ 0.1209 0.2025 ≤ a ¹ [4] ≤ 0.2026 0.2388 ≤ a ¹ [5] ≤ 0.2389 0.2025 ≤ a ¹ [6] ≤ 0.2026 0.1208 ≤ a ¹ [7] ≤ 0.1209 0.0467 ≤ a ¹ [8] ≤ 0.0468 0.0096 ≤ a ¹ [9] ≤ 0.0097 0.0008 ≤ a ¹ [10] ≤ 0.0009

In a preferred embodiment, the coefficients a ⁰ [ν] and a ¹ [ν] of the standard filter follow the relationship of the following Tables 8 and 9.

0.00375672984183 ≤ a ⁰ [0] ≤ 0.00375672984185 A ⁰ [1] | ≤0.00000000000010 -0.07159908629243 ≤ a ⁰ [2] ≤-0.07159908629241 A ⁰ [3] | ≤0.00000000000010 0.56743883685216 ≤ a ⁰ [4] ≤ 0.56743883685218 A ⁰ [5] | ≤0.00000000000010 0.56743883685216 ≤ a ⁰ [6] ≤ 0.56743883685218 A ⁰ [7] | ≤0.00000000000010 -0.07159908629243 ≤ a ⁰ [8] ≤-0.07159908629241 A ⁰ [9] ｜ ≤0.00000000000010 0.00375672984183 ≤ a ⁰ [10] ≤ 0.00375672984185

And

0.00087709635502 ≤ a ¹ [0] ≤ 0.00087709635504 0.00968961250933 ≤ a ¹ [1] ≤ 0.00968961250935 0.04670597747405 ≤ a ¹ [2] ≤ 0.04670597747407 0.12080166385304 ≤ a ¹ [3] ≤ 0.12080166385306 0.20257613284429 ≤ a ¹ [4] ≤ 0.20257613284431 0.23887175675671 ≤ a ¹ [5] ≤ 0.23887175675673 0.20257613284429 ≤ a ¹ [6] ≤ 0.20257613284431 0.12080166385304 ≤ a ¹ [7] ≤ 0.12080166385306 0.04670597747405 ≤ a ¹ [8] ≤ 0.04670597747407 0.00968961250933 ≤ a ¹ [9] ≤ 0.00968961250935 0.00087709635502 ≤ a ¹ [10] ≤ 0.00087709635504

의 값에 기초하는 값

을 가지고 다음의 수학식에 기초하는 중간 신호

의 값을 가진 상기 실수치 부대역 신호를 제공하도록 동작하고,In a preferred embodiment, the calculation unit 307 (760) is the real value subband signal

A value based on the value of

With an intermediate signal based on

Provide a real value subband signal having a value of

M is the exponent of the subband signal in the range from 0 to (K-1).

In a preferred embodiment, the devices 310 and 560 operate to receive a number of other real subband signals comprising (LK) subband signals, and the devices 310 and 560 are configured to receive the other plurality of threads. Operate to output a numerical subband signal, wherein L is a positive integer and L is equal to or greater than K.

In a preferred embodiment, the device 310 (560) is a device, wherein the integer L is specified to be 64.

In a preferred embodiment, the devices 310 and 560 further comprise a delayer 670 for delaying the plurality of real valued subband signals and passing the real valued subband signals in a delayed form.

In a preferred embodiment, the method according to the present invention comprises a method for processing a plurality of complex subband signals, each comprising a first complex subband signal and a second complex subband signal, respectively, to obtain a real value subband signal. Extracting a first imaginary part from the first complex subband signal; Extracting a second imaginary part from the second complex subband signal; Extracting a real part from a first, second, or third complex subband signal among the plurality of complex subband signals; Filtering the first imaginary part to obtain a first filtered imaginary part signal; Filtering the second imaginary part to obtain a second filtered imaginary part signal; Combining the first filtered imaginary part signal and the second filtered imaginary part signal to obtain an intermediate subband signal; And combining the real part with the intermediate subband signal to obtain the real value signal.

Depending on the requirements in the implementation of the characteristic method of the invention, the method of the invention can be implemented in hardware or software. The present invention can be implemented using a digital storage medium, and in particular, is implemented by interlocking a disk, such as a CD, a DVD, etc., which stores electronically readable control signals with a programmable computer system for performing the particular method of the present invention. Can be. Generally, the present invention is provided as a computer program product having a program code recorded on a mechanical read carrier, and when the computer program product is executed on a computer, the program code performs the characteristic method of the present invention. In other words, the characteristic method of the present invention may be a computer program having program code for performing at least one of the characteristic methods of the present invention when the computer program is executed on a computer.

Although the present invention has been disclosed through the description of specific embodiments, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention. . In addition, it should be understood that various changes may be made in the present invention by employing various embodiments without departing from the spirit of the invention as disclosed in or from the claims set forth below.

Claims (19)

1. An apparatus (210; 520) for processing a plurality of real valued subband signals including a first real valued subband signal and a second real valued subband signal to obtain a complex subband signal,
Filter the first real subband signal to obtain a first filtered subband signal, filter the second real subband signal to obtain a second filtered subband signal, and real real intermediate subband signals A multi-band filter (204; 401; 600) for supplying the real value intermediate subband signal by combining the first filtered subband signal and the second filtered subband signal to obtain? And
Combining the real value subband signal from the plurality of real value subband signals as a real part of the complex subband signal with a signal based on the intermediate subband signal as an imaginary part of the complex subband signal A calculation unit (215; 650) for providing the complex subband signal,
The apparatus (210; 520) is operative to assign an exponent m to each of the real subband signals in accordance with a center frequency associated with the real subband signal, such that the real subband signal with increasing exponent m is increased. Arranged according to a center frequency associated with the real subband signal, wherein the plurality of real subband signals comprise K real subband signals, where K is a positive integer and m is from 0 to (K−). The device which is an integer in the range up to 1). The method of claim 1,
The apparatus (210; 520) includes a delayer (203; 610) for delaying the real valued subband signal to provide the real valued subband signal in a delayed form to the calculator (215; 650). . The method according to any one of the preceding claims,
The apparatus (210; 520) comprises a gain adjuster (207; 640) for receiving the complex subband signal from the calculator (215; 650) and adjusting the value of the complex subband signal. The method according to any one of the preceding claims,
Wherein the plurality of real value subband signals are output by a real value QMF analysis bank (500). The method according to any one of the preceding claims,
A low pass filter characteristic, a high pass filter characteristic, or a band pass filter for the multi-band filter 204; 401; 600 to filter the first real value subband signal and to filter the second real value subband signal A device that operates by employing a characteristic. The method of claim 1,
The multiband filter (204; 401; 600) is operative to provide the exponent m to the real value intermediate subband signal in response to an exponent m associated with the first real value subband signal. The method of claim 6,
The multiband filter 204; 401; 600 is a real value subband from the plurality of real value subband signals associated with an exponent (m + 1) or an exponent (m-1) as the second real value subband signal. And operate to utilize the signal. The method according to claim 6 or 7,
The multi-band filter (204; 401; 600) filters a third real valued subband signal to obtain a third filtered subband signal, and obtains the first filtered subband signal, Combine the second filtered subband signal and the third filtered subband signal to provide the real value intermediate subband signal,
The exponent of the second real value subband signal is (m-m ') and the exponent of the third real value subband signal is (m + m'), or the exponent of the second real value subband signal is ( m + m ') and the exponent of the third real value subband signal is (m-m'),
M 'is a positive integer and m is the exponent of the first real value subband signal. The method of claim 8,
The multiband filter 204; 401; 600 is a real value intermediate subband for each real value subband signal as the first real value subband signal from a plurality of real value subband signals having an exponent mq (m). To provide a reverse signal,
The exponent of the second real value subband signal is m, and the exponent of the third real value subband signal is m + q (m). The method according to claim 1 or 9,
The multiband filter 204; 401; 600 is a value

Provide K intermediate real-valued subband signals with L, wherein n and m are positive integers and the K real-valued subband signals with the exponent m in a range from 0 to (K-1). For each of the following equations are defined,

Ν is an integer in the range from 0 to 10,

A ⁰ [ν] and a ¹ [ν] are coefficients of a standard filter,
Wherein each of the coefficients a ⁰ [ν] and a ¹ [ν] of the standard filter follows the relationship of the following Tables 10 and 11.
[Table 10]

And
[Table 11]

The method of claim 10,
Wherein the multi-band filter (204; 401; 600) is designed such that the coefficients a ⁰ [v] and a ¹ [v] of the standard filter follow the relationship of the following Tables 12 and 13.
[Table 12]

And
[Table 13]

The method according to claim 1 or 11, wherein
The calculation unit 215 (650) is the index m and the value

And provide K complex subband signals with the following equations,
K, n, and m are integers,
M is an integer in the range from 0 to (K-1),

remind

Denotes the value of the real value subband signal,

Denotes the value of the real value middle subband signal, and i denotes

Which represents an imaginary unit defined by. The method of claim 1 or 12,
The apparatus (210; 520) is operative to receive a plurality of other real value subband signals comprising (LK) real value subband signals, and the other plurality of real value subband signals as a real value subband signal. To provide,
Wherein L is a positive integer and L is greater than or equal to K. The method of claim 13,
And the device (210; 520) is designed such that the positive integer L is 64. The method according to claim 13 or 14,
The apparatus (210; 520) further includes a delay (202; 660) for delaying the real valued subband signal of the plurality of other real valued subband signals,
The apparatus (210; 520) is operative to provide the other plurality of real valued subband signals in a delayed form. An analysis filter bank 500 for processing the audio input signal into a plurality of real valued subband signals;
16. An apparatus according to any of claims 1 to 15, comprising: an apparatus (210; 520) for processing said plurality of real valued subband signals to obtain a complex subband signal;
A transformer (103) for receiving the complex subband signal and providing the complex subband signal in a modified form;
An apparatus 310 (560) for obtaining a real value subband signal; And
A synthesis filter bank (570) for processing said real valued subband signal into an audio output signal. 17. The method of claim 16,
The analysis filter bank 500 is designed such that the plurality of real value subband signals include L real number subband signals,
The L is a positive integer, and the device 210 (520) for processing the plurality of real valued subband signals is characterized in that the device (210; 520) has a plurality of complex subband signals and other multiple complex subbands. Designed to provide a signal,
The plurality of complex subband signals include K complex subband signals, the other plurality of real value subband signals include (LK) real value subband signals,
K is an integer in the range from 1 to L,
The transformer 103 is operative to transform the K complex subband signals of the plurality of complex subband signals to provide K complex subband signals in a modified form,
The system further includes a transformer 102 for modifying the other plurality of real valued subband signals and providing the other plurality of real valued subband signals in a modified form,
The apparatus (310; 560) includes a plurality of complex subband signals comprising K real number subband signals to obtain a final plurality of real value subband signals;
Designed to process the other complex subband signal comprising (LK) real subband signals,
The final plurality of real value subband signals comprise L real value subband signals,
The synthesis filter bank (570) is designed such that the final plurality of real valued subband signals are processed into an audio output signal. 1. A method of processing a plurality of real valued subband signals each comprising a first real valued subband signal and a second real valued subband signal to obtain a complex subband signal,
Filtering the first real subband signal to obtain a first filtered subband signal;
Filtering the second real subband signal to obtain a second filtered subband signal;
Combining the first filtered subband signal and the second filtered subband signal to obtain a real value intermediate subband signal; And
Combining a real value subband signal from the plurality of real value subband signals as a real part of the complex subband signal with a signal based on the intermediate subband signal as an imaginary part of the complex subband signal Including,
Assigning an exponent m to each of the real valued subband signals in accordance with the center frequency associated with the real valued subband signal, such that the real valued subband signal with increasing exponent m is at a center frequency associated with the real valued subband signal. And the assigning step, wherein the plurality of real value subband signals comprise K real number subband signals, where K is a positive integer and m is from 0 to (K-1). It is an integer in the range up to. A computer program for executing the method of claim 18 on a computer.

Images